DNA Nanodevices Friedrich C

Home , DNA computing, DNA nanotechnology, Molecular machine

reviews F. C. Simmel and W. U. Dittmer

Nanotechnology with DNA DNA Nanodevices Friedrich C. Simmel* and Wendy U. Dittmer

From the Contents

1. Introduction...... 285

2. Overview: DNA Nanotechnology...... 285

3. Prototypes of Nanomechanical DNA Devices...... 287

4. Unidirectional Motion Driven by DNA...... 292

5. DNA Devices Incorporating Functional Nucleic Acids...... 294

6. DNA Devices and Information Processing ...... 296

7. Outlook...... 297

Keywords: A DNA actuator, which switches from a relaxed, circular form to a stretched conformation. · DNA · molecular machines · molecular recognition · nanobiotechnology · nanodevices

284 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim DOI: 10.1002/smll.200400111 small 2005, 1, No. 3, 284 –299 DNA Nanodevices

The molecular recognition properties of DNA molecules combined with the distinct mechanical properties of single and double strands of DNA can be utilized for the construction of nanodevices, which can perform ever more tasks with possible applications ranging from nanoconstruction to intelligent drug delivery. With the help of DNA it is possible to construct machinelike devices that are capable of rotational motion, pulling and stretching, or even unidirectional motion. It is possible to devise autonomous nanodevices, to grab and release molecules, and also to perform simple information- processing tasks.

1. Introduction

Biological information is stored in the base sequences of nanostructures: First and foremost, the unique base-pairing DNA molecules. Highly specific base-pairing interactions interactions between complementary bases,[1, 2] second, the allow for replication of DNA and transcription of its infor- distinct polymer mechanical properties of single- and mation into RNA.[1] In other words, the specific interactions double-stranded DNA,[3] and finally, the electrostatic prop- between two nucleic acids are “programmed” into their se- erties of DNA and RNA as highly charged polyelectro- quences. Taken together with the polymer mechanical prop- lytes.[4] erties of DNA molecules, the notion of sequence-pro- Single-stranded DNA is a heteropolymer that consists of grammability has led researchers in nanoscience to think nucleotide units linked together via phosphodiester bonds about a utilization of DNA molecules for the construction (Figure 1). The nucleotide units themselves consist of a de- of artificial nanosystems. DNA can be used to build supra- oxyribose sugar unit to which one of four so-called “bases” molecular devices or as a template for materials synthesis. are attached at the 1’ carbon site. Two of the four bases— Quite recently, it has been shown that DNA can not only be adenine and guanine—are purines, and the other two—thy- utilized to build such static nanostructures, but also to con- mine and cytosine—are pyrimidines. In the famous Watson– struct simple, machinelike nanomechanical devices. This Crick (WC) base-pairing scheme, adenine can bind to thy- Review provides an overview of the DNA nanodevices real- mine through two hydrogen bonds, and guanine can bind to ized so far and the major current research themes in this cytosine via three such bonds. In RNA, uracil takes the role field. of thymine and the sugar unit is ribose. If the base sequen- In Section 2., a brief overview of DNA nanotechnology ces of two DNA strands are exactly complementary, that is, as a whole is given. The most important properties of DNA if for each base on one strand the corresponding WC part- molecules are introduced and examples are given in which ner is found on the other strand, the strands may bind to- DNA has been utilized for nanoconstruction and materials gether to form a double helix. Under typical buffer condi- synthesis. Section 3. surveys the prototypical DNA nanoma- tions (at least 100 mm salt concentration; neutral pH) a chines that have been realized so far. These are simple DNA duplex assumes its native “B” conformation. In this DNA constructs, which are capable of primitive movements form, the distance between two bases is 0.34 nm and the such as rotation or stretching. Section 4. discusses the impor- helix completes one turn for each 10.5 base pairs (bp). The tant problem of locomotion and recent advances in the con- diameter of the B-form duplex is 2 nm. For single-stranded struction of DNA motors. Section 5. introduces new devel- DNA, the mean distance between two bases is 0.43 nm, opments in which functional nucleic acids such as aptamers slightly greater than in the duplex. In RNA duplexes or in or ribozymes have been incorporated into DNA nanodevi- DNA–RNA hybrids, the “A” conformation is found with a ces. Section 6. gives an overview of attempts to process mo- diameter of 2.6 nm and 11 bp per turn. Under certain buffer lecular information using DNA structures and thereby con- conditions, and for special sequences, DNA may also under- trol the motion of DNA nanodevices. Finally, the outlook in go a transition to a left-handed helical conformation known Section 7. briefly discusses future directions and possible ap- as Z-form DNA, which has a diameter of 1.8 nm and 12 bp plications for DNA nanodevices. per turn.[2]

2. Overview: DNA Nanotechnology [*] Dr. F. C. Simmel, Dr. W. U. Dittmer Department of Physics and Center for Nanoscience 2.1. Important Properties of Nucleic Acids LMU Munich, Geschwister Scholl Platz 1, 80539 Munich (Germany) Several biophysical and biochemical aspects of DNA are Fax: (+49)89-2180-3182 particularly important for DNA-based nanodevices and E-mail: [email protected] small 2005, 1, No. 3, 284 –299 DOI: 10.1002/smll.200400111 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 285 reviews F. C. Simmel and W. U. Dittmer

erties of a given sequence can be calculated quite reliably using the wealth of thermodynamic data collected for DNA over the last decades,[6] for example, using the computer program HYTHER.[7, 8] Many of the DNA nanodevices discussed in the following Sections are driven by this hybridization energy. It is expected that the maximum force which may be generated by hybridization is on the order of 15 pN, a force which has also been measured by atomic force microscopy (AFM) in a variety of DNA-unzipping experiments.[9–11] If two strands of DNA do not match perfectly, they may bind to each other more weakly, or remain single-stranded, depending on the number of mismatches. DNA may also fold back upon itself to form “hairpin loops”. From a nanoconstruction point of view, this means that structure is determined by sequence, which opens up the possibility of pro- grammable nanoassembly. By choosing the appropriate DNA base sequences, one can design arbitrary networks consisting of single- and double-stranded sections. Due to the differing mechanical properties of single- and double-stranded DNA, self-assembled DNA structures can be thought of as networks consisting of relatively stiff Figure 1. Important features of DNA: a) Watson–Crick base pairs elements connected by flexible joints: The stiffness of a formed between adenine (A) and thymine (T), and guanine (G) and polymer can be characterized by its persistence length l , cytosine (C); b) in single-stranded DNA, nucleoside units (deoxyri- p bose + base) are linked by phosphodiester bonds; c) two comple- which is the correlation length for the tangent vector along [12,13] mentary strands of DNA form a double helix. Top-left: A symbolic rep- the polymer. On a length scale comparable to lp, a poly- resentation of a DNA duplex of 21 bp. Bottom-left: Looking down the mer can be regarded as a rigid rod, whereas for lengths

axis of the double helix. Right: Two helical turns of a duplex made much larger than lp, the polymer is flexible. It has been from 21 bp. found that the mechanical properties of double-stranded DNA can be well described by the wormlike chain (WLC) The stability of the double helix is determined by a vari- model, with a persistence length of typically 50 nm or ety of factors: the binding energy due to hydrogen bonds, 150 bp.[14–16] By contrast, single-stranded DNA has a much stacking interactions between neighboring base pairs, en- shorter persistence length, on the order of only 1 nm.[15,17,18] tropic contributions, and many others. The stability can be Thus, for the DNA structures of interest here with a length controlled by a variety of external parameters, such as salt on the order of 10–100 bases, duplex DNA really is a stiff concentration or temperature. At T=298 K and a monova- molecule, while single-stranded DNA is flexible. The elastic lent salt concentration of 1m, the mean value of the binding properties of DNA are sequence-dependent, which in some free energy between two bases in a double-helical context cases can be used to make fine adjustments. For instance, it 0 (i.e., containing stacking interactions) is DG298 75 meV has be shown that poly(dT) is much more flexible than (calculated from the thermodynamic parameters given else- poly(dA).[19] where[5, 6]). Due to their three hydrogen bonds, G–C pairs Under neutral buffer conditions the negative charges on are more stable than A–T pairs. The thermodynamic prop- the phosphate groups in the backbone render DNA a highly

Friedrich C. Simmel graduated in physics Wendy U. Dittmer graduated from the at the Ludwig-Maximilians-Universitt University of Toronto, Canada, with a Mnchen, Germany, in 1996 and ob- Bachelor of Science in Physical Chemis- tained his PhD in the group of Prof. Dr. try in 1997. She went on to obtain a PhD J. P. Kotthaus at the same university in in Physical Chemistry at the University of 1999. From 2000 until early 2002 he California, Berkeley, in 2002 with Prof. worked as an Alexander von Humboldt A. P. Alivisatos as an NSERC Scholar, (Feodor Lynen) postdoctoral fellow in the during which time she developed new group of Dr. B. Yurke at Bell Laboratories, nanocrystalline plastic solar cells. Cur- Lucent Technologies (Murray Hill, NJ, rently she is completing postdoctoral USA). Since 2002 he has led a junior re- work at the University of Munich as an search group with a focus on biomolecu- Alexander von Humboldt fellow in the lar nanoscience at Munich University, field of bionanotechnology.) which is funded by the Deutsche Forschungsgemeinschaft through its Emmy Noether program.

286 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim www.small-journal.com small 2005, 1, No. 3, 284 –299 DNA Nanodevices charged polymer.[4] This is quantified by the Oosawa–Man- over constructs,[36] DNA crossbar lattices,[37] or DNA nano- ning parameter x,[20, 21] which expresses the inverse line tubes.[38] charge density of a polyelectrolyte in units of the Bjerrum An interesting approach is the utilization of triangular 2 [39, 40] length lB =e /4pere0 kB T, where er is the permittivity of the building blocks for DNA nanoconstruction, since these solvent and kB is the Boltzmann constant. For DNA, x=4.2, are intrinsically rigid objects. One of the first demonstra- which is well above the critical value of x=1 above which tions of three-dimensional objects from DNA—the octahe- counterion condensation is expected to set in.[4] This has a dra—is also based on triangular subunits.[41] variety of important consequences: For templated materials Another innovative approach towards supramolecular synthesis, DNA may be used to localize positively charged construction is the utilization of RNA rather than DNA. precursor molecules along its backbone. Furthermore, de- Leontis and co-workers[42, 43] have devised a number of pending on the nature and concentration of the counterions, “RNA tectons” that specifically interact with each other two DNA duplexes may interact strongly with each other using non-WC interactions. In these RNA structures, loop electrostatically. The interaction is mediated by the counter- elements are recognized by specific binding pockets, which ion cloud and is usually repulsive but sometimes also attrac- are very similar to key–lock interactions between proteins. tive.[22,23] Under certain conditions, these effects can strongly influence the performance of DNA nanodevices, for example when a conformational change brings two DNA duplex 2.3. DNA Templating and Scaffolding units into close proximity. Apart from the advantageous properties of DNA, the To utilize DNA for the assembly of non-biological nano- many biochemical techniques available for its manipulation structures such as nanoelectronic circuits, it is necessary to and characterization make supramolecular construction with attach functional nanostructures to DNA or to chemically DNA particularly convenient. Many DNA-modifying en- modify DNA itself. The conductive properties of DNA itself zymes are available that can be used to cut DNA at specific are not sufficient for any application in a conventional elec- sites or to ligate two pieces of DNA together. Techniques tronic circuit. For this reason, a number of groups have such as polymerase chain reaction (PCR) or cloning can be modified DNA with electronically functional materials such utilized to make large numbers of copies of given DNA se- as metals,[44–49] semiconductor nanoparticles,[50, 51] conductive quences. But maybe the most important factor for DNA polymers,[52–54] or carbon nanotubes.[55,56] In most of these nanoconstruction is the capability of producing oligonucleo- cases, an ionic precursor such as a metal ion or metal com- tides of arbitrary sequence using automated synthesis meth- plex was attached electrostatically to the DNA backbone or ods. Therefore, the sequences necessary for a computer-de- was directly bound to the DNA bases. Subsequent chemical signed model can be readily translated into a physical real- steps such as reduction or polymerization led to the tem- world structure. plate-directed deposition of materials along the DNA molecule. In the case of carbon nanotubes, DNA was attached either covalently[55] or by utilization of antibody recognition 2.2. Supramolecular Construction with DNA events.[56] There is also a large body of work in which DNA has been used to organize colloidal metallic and semicon- Since the early 1990s, the unique molecular recognition ductor nanoparticles[57–60] or proteins[61] into crystals and net- properties of complementary strands of DNA and the me- works. chanical properties of single-stranded and duplex DNA A different approach is the utilization of DNA recogni- mentioned in the previous Section have been utilized for tion for organic synthesis. Here reactive chemical com- the construction of a variety of geometric objects and two- pounds are attached to complementary strands of DNA. dimensional lattices. There are a number of excellent re- Duplex formation between the complementary strands views on these structures,[24–26] and we just mention a few of brings the compounds into close proximity, so that they can the hallmarks in this field. The potential of DNA was first react with each other. A thorough review of applications of utilized by Seemans group in the realization of a supra- this concept has been published recently.[62] molecular structure with the topology of a cube[27] and later of a truncated octahedron,[28] or of DNA catenanes such as Borromean rings.[29] The construction of two-dimensional 3. Prototypes of Nanomechanical DNA Devices networks of DNA from four-way junctions was initially compromised by the relatively large angular flexibility of DNA can not only be used to build supramolecular the branch points of the junctions.[30–32] This problem was re- structures or act as a template for materials synthesis; it can solved with the introduction of the “double-crossover” con- also be utilized to produce nanoscale movements. These structs in which two four-way junctions were joined to form movements are usually based on conformational changes, an inherently more rigid object.[33] From these “DX tiles”, which are triggered by changes in buffer composition or by Winfree et al. constructed the first two-dimensional DNA hybridization between complementary strands of DNA. The lattices.[34] The persistence length of DX structures was later conformational changes are transitions between single- shown to be twice that of duplex DNA.[35] Recently, other stranded DNA and duplex DNA or between unusual and complex structures have been realized, such as triple-cross- conventional DNA conformations. In the following, we dif- ferentiate between DNA devices driven by small molecules small 2005, 1, No. 3, 284 –299 www.small-journal.com 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 287 reviews F. C. Simmel and W. U. Dittmer

and DNA devices driven exclusively by hybridization and proximately 5.6 nm. The transition from the B-form to the branch migration. Z-form leads to a change in FRET efficiency from 20% to 5% due to the larger distance between the dyes in the Z conformation of the device. Using the FRET technique, it 3.1. DNA Devices Fuelled by Changes in Buffer Composition could be shown that the B–Z device indeed can be cyclically switched between its two conformations and thus represents The first DNA nanomechanical device demonstrated a very simple example of a molecular machine based on was based on the B–Z transition of DNA.[63] In the B–Z DNA. device, two double crossover structures are connected by a A different approach towards nanoscale motion was [67] DNA strand containing the sequence d(CG)10 (where cyto- taken by Niemeyer et al.: In their device, magnesium-ion- sine residues are methylated at the C5 position). This se- induced DNA supercoiling was utilized to produce nano- quence is particularly prone to undergoing a B–Z transition. scale movements that could be characterized using AFM. In The transition can be triggered by a change in the concen- networks of dsDNA connected by biotin–streptavidin link- 3+ tration of cobalt hexammine ([Co(NH3)6] ) from 0 to ers, two neighboring DNA duplexes may condense into a m 0.25 m . The d(CG)10 section changes its helicity from single supercoiled structure in the presence of multivalent right-handed to left-handed; this corresponds to a change in ions such as Mg2+, which results in a change of network twist of 3.5 helix turns. From this, an overall change in the connectivity. In principle, it should be possible to reversibly conformation of the device results in which the two DX sec- switch between the conformations of such a network as the tions are rotated with respect to each other by one half-turn supercoiling transition itself is reversible. (see Figure 2). The motion of the B–Z device can be moni- A concept termed “duplex pinching” by Fahlman et al.[68] is based on the guanine quadruplex conformation of DNA (Figure 3a).[69] In the G quadruplex, four guanine resi-

Figure 2. The B–Z device is made from three strands of DNA. One strand (dark and light gray) forms two double-crossover (DX) arms Figure 3. Nonstandard interactions between DNA bases: a) A guanine with two other strands (gray). The hinge (light gray) connecting the quartet formed by cyclic hydrogen bonding between four guanine arms is a d(CG)10 sequence which undergoes a B–Z transition upon bases; b) a base pair between a protonated (C+) and an unproton- addition of cobalt hexammine. This transition rotates the DX arms ated (C) cytosine. with respect to each other (reproduced with permission from Ref. [63]) dues can form a total of eight hydrogen bonds with each other in a cyclic arrangement. Under certain conditions, the tored in fluorescence resonance energy transfer (FRET) ex- strength of this interaction is comparable to conventional periments.[64–66] FRET occurs between a fluorescent dye (the WC base-pairing. G-quadruplex formation is often found in “donor”) and another chromophore (the “acceptor”) if the G-rich sequences and can occur not only intramolecularly, absorption band of the acceptor overlaps with the emission but also intermolecularly, where two or four single strands

band of the donor fluorophore. The efficiency ET of this of DNA are held together by G quartets. For their devices, process varies as Fahlman et al. made use of the fact that the G-quadruplex conformation is stabilized by the presence of potassium or FðDÞFðDAÞ 1 strontium ions. They constructed a DNA duplex with six in- ET ¼ ¼ 6 ð1Þ FðDÞ 1 þðR=R0Þ ternal G–G mismatches connected by a short hinge section consisting of three A–T base pairs. Upon addition of Sr2+ where F(DA) and F(D) denote the fluorescence intensi- ions, which strongly stabilize G-quartet formation, the ties of the donor in the presence or absence of the acceptor. duplex assumes a “pinched” conformation in which the The distance R0 at which FRET efficiency is 50% is called G–G mismatches form intramolecular G quartets. This tran- the Fçrster distance; R0 usually lies between 2 and 6 nm. sition can be reversed by the addition of the chelator ethyl- For this reason, FRET is a convenient tool to characterize enediamine tetraacetic acid (EDTA) which binds the stron- nanoscale movements and is a frequently applied technique tium ions. The authors proposed that duplex pinching may in the field of DNA nanodevices. The B–Z device was char- be used to introduce contractile elements into DNA supra- acterized using fluorescein as the FRET donor and Cy3 as molecular structures. the acceptor. For this FRET pair, the Fçrster distance is ap-

288 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim www.small-journal.com small 2005, 1, No. 3, 284 –299 DNA Nanodevices

A variety of other DNA devices have been constructed are protonated, a duplex–triplex transition can be induced that are fuelled by protons rather than by metal ions.[70,71] by changing the pH value. Mao and co-workers recently in- Proton-fuelled devices utilize the fact that nucleic acid bases serted an appropriate sequence into a DNA nanomechani- can be protonated at low pH values. For instance, adenosine cal device, the conformation of which could be switched by can be protonated at the N1 nitrogen position (pKa =3.5) inducing a duplex–triplex transition in this “motor sec- [71] and cytidine can be protonated at the N3 position (pKa = tion”. A simpler DNA construct utilizing the duplex–tri- 4.2).[2] These protonated bases can form additional non-WC plex transition has already been described elsewhere.[75] base pairs, for example, C–C+ (Figure 3b). The DNA device realized by Liu and Balasubramanian[70] is based on an unusual DNA conformation called the “i-motif”, which 3.2. Devices Driven by DNA Hybridization and Branch occurs in long stretches of cytosine residues with a transition Migration at around pH 6.5.[72–74] Below this value, a C-rich DNA strand X assumes an intramolecular quadruplex conforma- A different operational principle for DNA devices was tion in which C–C+ base pairs occur in a staggered ar- introduced by Yurke et al. with the “DNA tweezers”.[76] In rangement (see Figure 4). At higher pH values, strand X this concept, a 40-base-long strand Q is hybridized with two 42-nucleotide (42 nt) strands S1 and S2 over a length of 18 bp each. This results in a supramolecular structure, as de- picted in Figure 5, in which two 18-bp-long duplex “arms” are connected by a 4-nt-long single-stranded flexible hinge.

Figure 4. A proton-fuelled device based on the i-motif. In the closed state at low pH values the cytosine residues in strand X form C–C+ base-pairs. These stack up on top of each other to form the “i-motif”. At higher pH values, cytosine is deprotonated and X can base-pair with the complementary strand Y to form the open state. Reproduced with permission from Ref. [70]. can hybridize with the complementary strand Y. Therefore the device can be cycled between an extended duplex state and a closed quadruplex state by the subsequent addition of Figure 5. DNA tweezers. In the “open” state the tweezers structure is hydrogen and hydroxide ions. By attaching a fluorescent formed by three strands S1, S2, and Q. As indicated, the “fuel” dye at the 5’ end and a quencher group at the 3’ end, the strand F can hybridize partly with the single-stranded section of S1 conformational changes of the device can be monitored by and partly with S2. This closes the tweezers structure. In the “closed” state, F still has an unhybridized single-stranded section changes in fluorescence intensity as the quenching efficiency (the “toehold”), where the complementary “removal” strand R can is strongly dependent on the distance between the chromo- attach. F is removed from the tweezers by R via branch migration phores, similar to the FRET effect introduced above. From (see Figure 6) and restores the open state. Two fluorescent dyes are the free-energy difference of the two states and the change indicated (black and gray dots), which are used for FRET characteriza- in distance between the endpoints of strand X, the authors tion of the device (adapted from Ref. [76]). inferred an average closing force of 11.2 pN. As “waste products” only NaCl and H2O are produced. Two 24-nt-long sections of S1 and S2 remain unhybridized Another device driven by protons utilizes DNA-triplex and form single-stranded extensions of the arms. To drive formation. Triplex formation can occur when an oligonu- the machine into a closed conformation a fourth, 56-nt-long cleotide binds to the major groove of a homopurine–homo- DNA strand termed the “fuel strand” F is added of which pyrimidine duplex.[69] There are two possible orientations 24 bases are complementary to the unhybridized section of for the third strand, namely, parallel or antiparallel to the S1 and another 24 bases are complementary to the exten- homopurine stretch. One of the common triplex motifs in- sion of S2. The hybridization of F with these device sections volves C–G:C+ triplets, where the third strand is oriented pulls the arms of the tweezers structure together. The re- in parallel with the homopurine sequence. As this triplex maining eight unhybridized bases of F serve as a point of at- can only form when the cytosine residues in the third strand tachment for another DNA strand R (the “removal small 2005, 1, No. 3, 284 –299 www.small-journal.com 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 289 reviews F. C. Simmel and W. U. Dittmer

Figure 6. Branch migration: Apart from the gray “toehold” section, strands A1 and A2 are both complementary to strand C. The branch point can perform a random walk along strand C. Due to the toehold section, A1 cannot be completely displaced from C by A2 and finally “wins” the competition.

strand”), which can displace F from the device by a process called “branch migration”. In three-stranded branch migration, two homologous DNA strands A1 and A2 try to bind to one complementary strand C (see Figure 6). At the branch point where the two competing strands meet, thermal fluctuations may break a base pair, for example, between A1 and C, which gives A2 the opportunity to make a base pair with C. Such processes lead to a random motion of the branch point along C. In the case of short sequences C, one of the competing strands A1 or A2 will finally bind with its full length and completely Figure 7. Stretching motion of the DNA actuator: The actuator is simi- displace the other strand from C. The four-stranded version lar to the DNA tweezers, but with the two arms connected by a of this process is well known from genetic recombination in single-stranded loop. Hybridization of the loop with fuel strand F which two homologous DNA duplexes “cross over” and ex- stretches the device. The relaxed state can be restored by branch change strands.[1] Three-stranded branch migration with an migration using the removal strand R. A different kind of fuel strand average step time of approximately 10–20 ms[77,78] is consider- can be used to close the structure analogously to the DNA tweezers ably faster than its four-stranded counterpart (step times of (adapted from F. C. Simmel, B. Yurke, Phys. Rev. E 2000, 63, 035320). one to several hundreds of ms[78]). For short stretches of DNA, strand removal occurs relatively fast. The removal of a 40-base-long DNA molecule takes about 20–80 ms.[79, 80] Another variation of the tweezers was termed “DNA In the case of the tweezers, branch migration leads to scissors”.[83] In this case, two sets of tweezers structures are the formation of a “waste” duplex F–R and returns the S1- joined at their hinges by short carbon linkers. The motion of Q-S2 structure of the tweezers to their starting configura- one set of tweezers is transduced to the other part, resulting tion. By the alternate addition of fuel and removal strands, in a scissors-like movement. the device can be cycled through its “open” and “closed” A more complex device was constructed by Seeman and states many times. As described for the DNA devices in the co-workers based on multiple crossover motifs.[84] Using previous Section, the operation of the tweezers can be branch migration, a DNA structure can be switched be- monitored by FRET between fluorophores attached to the tween a “paranemic crossover” conformation and a “juxta- ends of the arms of the device. posed” configuration. The paranemic crossover motif is a Quite generally, strand displacement by branch migra- four-stranded DNA structure which can be thought of as tion provides a tool by which DNA hybridization can be consisting of two adjacent double helices of which strands employed for DNA devices in a reversible manner. Duplex of the same polarity are exchanged at every possible site formation can be used to pull molecules together or stiffen (Figure 8a).[85] If parts of this motif are removed and re- the connection between two nodes in a network. These placed by DNA sections without crossover, molecules such

structural transitions can be reversed by adding recognition as the “juxtaposed” JX2 structure (Figure 8a) result in tags to the DNA effector strands, which are recognized by which two helices are rotated by 1808 with respect to the removal strands and from where branch migration can corresponding section in the “paranemic crossover” PX

begin. This principle has been utilized recently in a variety structure. In the PX–JX2 device, the central section of a con- of other DNA nanodevices. struct based on the PX motif is defined with DNA mole- In a DNA “actuator”, the single-stranded extensions of cules, which can be removed by branch migration (Fig- the arms of the DNA tweezers are joined to form a closed ure 8b). After removal, these molecules can be replaced by [81,82] “motor section”. This allows the device to perform both DNA strands that switch the device to the JX2 configura- pulling and stretching movements (see Figure 7). tion. Again using branch migration, the device can be switched back to the original PX structure. As shown by

290 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim www.small-journal.com small 2005, 1, No. 3, 284 –299 DNA Nanodevices

3.3. The Problem of Free-Running Devices and Waste Products

The devices described so far are all driven by a similar principle: An external operator adds some sort of chemical fuel (ions, protons, DNA strands) that drives the system out of equilibrium, that is, under the new conditions the devices are not in a thermodynamically stable state. The system then decays to a “new” equilibrium, which results in a conformational change accompanied by some sort of motion. This process can be reversed by adding a different kind of fuel (buffer exchange, chelators, hydroxide ions, complementary DNA strands), which neutralizes the effect of the first fuel chemical. Again, the system decays to the new equilibrium state which is, almost, the same as the old one. There are a variety of problems with this approach. First of all, the devices described so far are operated in closed systems, that is, at constant temperature and pressure with no Figure 8. The paranemic crossover device: a) Two double strands of material exchange with the environment except when the DNA cross over at every possible site to form a paranemic crossover (PX) structure. Removing the inner two crossovers juxtaposes the fuel is added. This means that the waste products resulting from the reaction between fuel chemicals remain in so- helices D and C with respect to each other (JX2); b) from this PX–JX2 transition, a reversibly switchable device can be constructed. The lution. Therefore, of course, the system after one operation innermost double crossover of the PX structure can be resolved by cycle is not exactly the same as the starting system. In the removing the green strands by branch migration (1). Insertion of the case of proton-driven devices, the waste products typically violet strands leads to the formation of the JX2 structure (2). JX2 can are NaCl and H2O. This leads to a continuous dilution of be switched back to PX by another branch-migration and strand- the devices and also demands a continuous increase in the insertion process (3,4). Reproduced with permission from Ref. [84]. amount of fuel added. For DNA-driven devices, the problem arises that fuel Seeman and co-workers,[84] the switching between the two and removal strands have to be chosen in a complementary states of the device could be impressively visualized with fashion. When working with stoichiometric quantities, each AFM. Yan and co-workers recently showed that even the reaction cycle produces an amount of waste duplex DNA topology of large supramolecular DNA networks could be equal to the total amount of device strands. A practical switched employing a similar mechanism.[86] problem related to this is to add the strands in exact stoichi- Other devices utilizing branch migration are based on ometry. If the fuel and removal strands are not added in ex- unusual DNA tertiary structures such as the G quartets al- actly equal amounts, in each cycle some of the strands will ready encountered in Section 3.1. In the experiments descri- be left over and immediately react with fuel or removal bed by Li and Tan[87] and by Alberti and Mergny,[88] a strands in the next cycle. This quickly deteriorates the single-stranded DNA structure formed by stacked intramo- device performance and ultimately brings it to a halt. If one lecular G quadruplexes is switched to a stretched duplex tries to overcompensate stoichiometric errors in each cycle structure by the addition of the complementary DNA fuel one faces the problem of an exponential build-up of waste strand. This transition can be reversed by removing the fuel strands and an exponentially growing need for fuel. A more strand using branch migration as described above. fundamental problem, of course, is that the accumulation of Among the great advantages of using DNA as a fuel waste will ultimately poison the system. Increasing amounts rather than a small molecule is its sequence specificity— of waste products make it more difficult to drive the system DNA not only drives the motion, it also serves as an address out of equilibrium as their presence favors the “backward” label for a device. In a situation with several distinct DNA reaction. For most practical purposes so far, where devices devices, specific devices can be operated with specific fuel have only been cycled on the order of 5–10 times, this is not molecules. Recently this notion was particularly emphasized a severe problem. Waste duplexes with a length of 20–40 bp by Shin and Pierce,[89] who have specifically addressed are extremely stable (the half life for the decay of an 18 bp single-stranded DNA labels attached to a double-stranded duplex at 208C and 1m NaCl is on the order of 109 years!) DNA scaffold. DNA or RNA strands can also be seen as a and therefore a back-reaction is not expected under normal specific molecular signal to which a DNA device can re- experimental conditions. In the experiments by Yan and co- spond by mechanical action (examples for this will be dis- workers,[84, 86] the waste problem has been tackled by using cussed in Section 6.). biotinylated fuel strands, which can be removed with streptavidin-coated magnetic beads after each reaction cycle. The concept of alternating the addition of fuel and neu- tralizing chemicals also makes it difficult to construct continuously running devices that operate for several cycles without interference by an external operator. In the case of small 2005, 1, No. 3, 284 –299 www.small-journal.com 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 291 reviews F. C. Simmel and W. U. Dittmer

the DNA tweezers and similar devices, fuel and removal devices will require techniques for single-molecule observa- strands cannot be added simultaneously as they would im- tion as they are common for the characterization of biologi- mediately react with each other rather than driving the cal molecular motors.[13] device through its states. One possible solution for this The problems discussed in this Section also point to- problem is based on the control of hybridization kinetics be- wards an important difference between the artificial devices tween complementary strands.[90] As indicated in Figure 9a, realized so far and biological motor proteins: Biological systems are open systems that operate under non-equilibrium conditions. The fuel consumed by molecular motors (such as ATP) is a metastable chemical, which is continuously sup- plied as a result of energy-producing catabolic reactions to which motor action (in general, in the form of energy-con- suming reactions), is ultimately coupled.

4. Unidirectional Motion Driven by DNA

One of the most challenging goals in the field of artificial molecular devices is the construction of a molecular machine with similar capabilities to a biological molecular motor. In biology, forces are generated in a variety of ways, Figure 9. a) “Protection” strand P forces F into a loop conformation. for example, by polymerization of cytoskeletal filaments or In this conformation hybridization with the complementary strand F¯ by conformational changes of motor proteins such as is slowed down considerably; b) in contrast, the short catalyst myosin or kinesin.[13,91–93] These motor molecules are respon- strand C can open the loop structure (b1). F¯ can attach to the sible for most of the motion exhibited by living organisms. opened loop structure and displace strands P and C from F (b2). C They drive cilia and flagella, are responsible for muscle con- continuously switches between a single-stranded conformation and a stretched duplex conformation (in F-P-C), until all of the fuel FP + F¯ traction and also actively transport substances within [13,94] is used up (adapted from Ref. [90]). cells. A large variety of other processes in living cells are carried out by complexes of proteins that can be viewed as sophisticated assembly machines.[1] Many of the features the hybridization between two complementary strands F exhibited by these biological examples would also be desira- and F¯ can be slowed down considerably by forcing the ble in an artificial context, for instance, to aid in nanoscale strand F into a loop conformation using a protection strand assembly processes or to provide micro- and nanostructures P. Hybridization can be accelerated by using a “catalyst” with the capability of movement. strand C, which can open the loop structure (Figure 9b). Biological motors like the myosins, kinesins, or dyneins The intermediate structure F-P-C can be attacked by F¯, are complex proteins that walk on supramolecular tracks which leads to the formation of the waste products F–F¯ and such as actin filaments or microtubuli.[13,94] Even though it is P. Strand C is then available to open another F–P structure. hard to imagine how such a sophisticated machinery could Such a catalyst strand C could be incorporated into a DNA be built on the basis of DNA alone, the general features of device, which would then be continuously driven through its these motors can serve as a guideline for the construction of states by the consumption of the metastable “fuel pair” F–P artificial DNA motors. In analogy to the biological motors, and F¯. The fuel molecules could be added simultaneously as the recent DNA “walker” systems (which will be discussed a reaction with the DNA device would be much faster than later) consist of tracks made of DNA that have single- a direct reaction. This approach is analogous to the enzy- stranded “binding sites” for a DNA walker. The walker can matic action of biological molecular motors and machines. be just a single DNA strand or a slightly more complex In biological systems, energy is stored in metastable fuel DNA structure, which can be translocated unidirectionally molecules such as ATP. The breakdown of ATP to ADP along the DNA track using various techniques. Directionali- and phosphate is catalyzed by molecular motors, and the ty is introduced by using unique sequences for the binding energy gained from the reaction is used to produce motion. sites, that is, essentially by making an asymmetric track. So Other approaches towards free-running systems based on far, this has resulted only in very simple devices capable of enzymes or on ribozyme action are discussed in Sections 4.1. a rudimentary form of locomotion. However, these are im- and 5.3. portant “first steps” in the right direction. It is clear that the realization of self-running devices will also require the adaptation of new experimental techniques for research on DNA nanodevices. So far, the “clocked” op- 4.1. Walking with the Help of Enzymes eration of DNA devices greatly simplifies their characterization as their function can be assessed in bulk experiments. An ingenious approach to achieve autonomous unidirec- All the devices are driven out of equilibrium at the same tional motion using DNA and three DNA-modifying en- time and collectively (but incoherently) change to the con- zymes was employed by Yin et al.,[95] who created a double- formation of their next state. Unsynchronized free-running stranded DNA “track” with three anchorages A, B, and C,

292 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim www.small-journal.com small 2005, 1, No. 3, 284 –299 DNA Nanodevices which are connected to the track with a 4 nt hinge (see 4.2. DNA Walkers and Gears Figure 10). Each anchorage has a 3 nt unhybridized section at its end. The DNA walker itself consists of a total of six In contrast to the device discussed in the previous Sec- nucleotides that reside on the ends of the two strands of an- tion, the following devices do not need enzymes for their chorage A in the starting configuration (denoted A*). An- operation. However, they are not autonomous and have to chorage A* can hybridize with the sticky end of anchorage be driven by the addition of fuel strands. B, then T4 ligase is used to covalently link the anchorages A very simple walker related to the rewritable DNA to form the structure A*B. The base sequences of A and B memory mentioned in Section 3.2. was recently demonstrated by Shin and Pierce.[96] A double-stranded DNA scaffold with single-stranded DNA labels attached to it served as a track for the walker—similar to the scaffold the authors described elsewhere.[89] The walker itself is a DNA duplex with two single-stranded extensions. As described in Figure 11, specific DNA strands are used to connect the single-stranded extensions to the labels on the track. The connector strands are equipped with single-stranded toehold sections and can be displaced from the device by their complementary strand via branch migration. The free-DNA “foot” can then be connected to the next free-label strand on the track. This can be repeated several times with the appropriate connector and removal strands to move the walker to arbitrary addresses on the track.

Figure 10. An enzyme-assisted DNA walker: The six nucleotides marked in red are transported from anchorage A to anchorage C in steps I–IV. This is achieved by joining neighboring anchorages with T4 ligase and cleaving with two different restriction endonucleases (PflM I and BstAP I). The operation principle of the device is described in more detail in the text (reproduced with permission from Ref. [95]). are chosen in such a way that A*B contains a site that is recognized by the restriction enzyme PflM I. After the cutting of A*B by the enzyme, the six walker nucleotides reside on anchorage B (i.e., B*). The new sticky end can now base-pair with anchorage C, and then B* and C are again covalently linked by T4 ligase. The duplex B*C now contains the recognition site of the endonuclease BstAP I. After cutting by the enzyme the six nucleotides of the walker are attached to anchorage C. Overall, the six nucleo- Figure 11. A DNA walker based on hybridization and branch migra- tides originally residing on the end of anchorage A have tion: a) The walker consists of duplex DNA with two single-stranded been translated to C by the action of T4 ligase, PflM I, and “feet”. The track is duplex DNA with single-stranded extensions as binding sites for the walker; b and c) the feet of the walker can be BstAP I without external interference. Ligation consumes attached to the track using linker molecules A1 and A2; d) a linker ATP as fuel and constitutes the irreversible step in the can be displaced from the track by a removal strand (D1). The foot is device operation. The direction of the motion is determined then free to bind to the next binding site on the track using another by the restriction sites and the walker cannot step back. linker molecule (reproduced with permission from Ref. [96]). small 2005, 1, No. 3, 284 –299 www.small-journal.com 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 293 reviews F. C. Simmel and W. U. Dittmer

A very similar principle, but with a more complex ar- gives kinesin and myosin proteins their directionality.[13] The rangement, was adopted by Sherman and Seeman.[97] Here DNA walkers simply detach from the track by some mecha- the walker (a “biped”) consists of two DNA duplexes nism and find their next binding site diffusively. A power joined by flexible single-stranded linkers. This biped device stroke may not be necessary for many applications, but it is can be translocated along a track that consists of a triple unclear how well diffusive DNA motors will work under a crossover (TX) molecule containing single-stranded address load. labels. The duplexes of the biped also contain single-stranded extensions and can be attached to the addresses on the track via DNA linker molecules. Again, these connections 5. DNA Devices Incorporating Functional Nucleic can be removed by branch migration and the free foot can Acids be connected to the next “foothold”. A different kind of unidirectional motion was recently 5.1. Aptamers and Ribozymes realized in the “molecular gears” system by Tian and Mao.[98] In this system, two DNA “circles” roll against each The role of DNA in the previously mentioned machines other driven by a mechanism based on hybridization and is to provide a structural basis and the chosen base sequence branch migration. The two circles consist of a closed single determines the switchable elements that enable reversible strand C, to which three other strands are hybridized. These motion. Thus, the DNA is relatively passive. To progress to strands contain flexible hinges made of thymine residues the next stage of complexity, beyond just movements result- and single-stranded extensions (termed “teeth”). The two ing from configurational changes, DNA machines should be circles can be connected by DNA linker strands, which are able to perform functional tasks. This requires the incorpo- partly complementary to one “tooth” of one circle and an- ration of functional and active units within these devices. other tooth of the other circle. Due to the flexible hinges in Functional nucleic acids are found in aptamers and ribo- the construction, two circles can be actually connected by zymes.[99–101] These are specific sequences of generally two linker strands simultaneously. As in the case of the single-stranded RNA and DNA, which adopt a tertiary walkers, the linker strands contain single-stranded overhang structure capable of a variety of activities including ligand sections to which removal strands can attach and displace recognition[102–104] and the catalysis of a number of chemical the linker strands from the device, thus leaving the circles reactions, among them cleavage,[105–107] ligation,[108, 109] and with only one connection. By the alternate addition of peptide-bond formation.[110] linker and removal strands in the correct order, the two cir- Although natural ribozymes can be found, the vast ma- cles can be made to roll against each other in one direction. jority have been artificially created through in vitro evolu- It is important to note that only the enzyme-driven devi- tionary selection techniques. In the systematic evolution of ces can work autonomously so far. All the other walkers ligands by exponential amplification (SELEX),[99] DNA or have to be driven from outside by a rather awkward proce- RNA sequences capable of specific and high-affinity binding dure in which they are forced to take one step after another to proteins or small molecules are isolated from a pool of and each binding site has to be specifically addressed. In over 1015 chemically synthesized sequences, which have these cases, an external operator actually has to keep track been partitioned and amplified based on their ability to of where the walker currently stands in order to move it in bind to the target. After 5–10 rounds of selection, the pool one direction. To displace a device over a longer distance, is highly enriched with sequences that have the desired many cycles of binding and unbinding have to be externally ligand-binding property. Modifications of this process also controlled. allow for the selection of sequences capable of catalyzing re- With the exception of the triple-crossover track in the actions such as cleavage and the ligation of nucleic acids.[101] device described by Sherman and Seeman,[97] the tracks As was anticipated in a review by Niemeyer and used for the walkers are rather floppy. This is not a problem Adler,[111] one of the next logical steps in the development at the moment as all the prototype walkers only make one of DNA-based molecular machines was to include function- step or two. To achieve locomotion over longer distances, al nucleic acids into DNA nanodevices. By combining ap- however, the track would need to be more stiff. Floppy tamer and ribozyme sequences with the structural and tracks also would lead to extensive crosslinking within a switchable elements in DNA nanomachines, devices capable track or between several tracks, as a DNA walker could of performing useful tasks can be constructed. One such easily bind to a distant binding site. The DNA tracks used functional nanomachine containing an aptamer sequence so far are basically double-stranded DNA with nicks at the was developed to bind and release a specific protein upon positions of the labels, that is, their persistence length is ex- instruction. The catalytic property of ribozymes has also pected to be less than that of duplex DNA. In comparison, been taken advantage of to allow a DNA machine to con- biological “tracks” such as F-actin or microtubuli have per- tinuously switch between two states. sistence lengths of 15 mm and 6 mm.[13] Possible solutions to this problem are to use stiffer DNA structures as tracks (such as the TX molecules) or to immobilize the DNA track 5.2. Aptamer-Based DNA Devices in an appropriate way. Finally, none of the devices are based on a conforma- Although nucleic acids have only four different subunits tional transition comparable to the “power stroke” that and thus, compared to proteins, a limited number of chemi-

294 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim www.small-journal.com small 2005, 1, No. 3, 284 –299 DNA Nanodevices cal moieties for binding, aptamers can bind a surprisingly 5.3. Autonomous Machines based on Nucleic Acid Enzymes broad range of proteins and small molecules with affinity comparable to that of antibodies to their antigen. This func- Hybridization-driven machines rely on the external addi- tionality of aptamers has been used in the construction of a tion of single-strand DNA at well-defined stages to bring machine that can controllably bind and release a single pro- about motion. One major aim for nanomachines is the de- tein molecule. A “nanohand” was created by combining a velopment of strategies for autonomous operation in which DNA aptamer selected to bind the human blood-clotting the machines function free of human intervention and ulti- protein, thrombin, with the operating principles of DNA- mately independently act based on decisions made in re- based nanomachines.[112] The 15 nt aptamer sequence 5’- sponse to stimuli (see also Sections 3.3. and 6.). One possi- GGTTGGTGTGGTTGG-3’ folds into a secondary struc- ble realization of autonomy was demonstrated by Chen ture consisting of two G-quartets, in the presence of potassi- et al. in a machine consisting of a DNA actuator[81] modified um ions (Figure 12). The machine in its standard state binds with an RNA-cleaving DNA enzyme.[113] In its standard thrombin. It can be instructed to release the protein by the state, the DNA enzyme is in a compact form and results in addition of single-stranded DNA containing a sequence that a closed-state for the actuator (Figure 13). In the presence of an RNA substrate (actually, a DNA/RNA chimera), the

Figure 13. An autonomous device incorporating a DNA enzyme: a) The device consists of two duplex arms formed by the strands F A switchable aptamer device: In the folded state (1) the Figure 12. and E. Strand E contains the sequence of an RNA-cleaving DNA aptamer device A can bind to the protein thrombin (T). In (2) the enzyme. When the chimeric RNA/DNA substrate S binds to strand E, opening strand Q attaches to the 12 nt long toehold section 5’- the arms of the device are opened; b) substrate S is cleaved by the TAAGTTCATCTC-3 of the device and displaces the protein. In the ’ enzyme into shorter strands S1 and S2; c) the duplexes with S1 and stretched duplex conformation (3) the device does not bind to throm- S2 are less stable and the strands S1 and S2 dissociate from strand bin. The removal strand R can attach to a second toehold section in E. The device then returns to its collapsed initial state. This operation strand Q and displace Q from A by branch migration. Strand A then cycle is repeated autonomously until all substrate is use up (repro- refolds and binds the protein thrombin again (reproduced from duced with permission from Ref. [113]). Ref. [112]). is complementary to a portion of the aptamer sequence; enzyme extends to its catalytically active structure as it this results in an unfolding of the aptamer to form a duplex, forms a duplex with the substrate and the actuator assumes which is unable to bind the protein. The protein can be its open configuration. Cleavage of the substrate causes a bound once again by the addition of another DNA strand, weakening of the hybridization forces and the cleaved sub- fully complementary to the previously added single-strand strate is released, which leads to the collapse of the enzyme DNA, which permits the aptamer machine to be released and thus an automatic closure of the actuator. As long as and return to its protein-binding quadruplex form. The op- the substrate is present, the machine can undergo several eration of this machine can be monitored using FRET with cycles of operation, in this case 20 in 30 min, independent of dyes positioned as shown in Figure 12. The release of pro- external influence. When using a non-cleavable DNA sub- tein by the machine can also be directly followed with fluo- strate, the device can also be deliberately forced into a stal- rescence anisotropy experiments. In principle, such a nano- led configuration.[114] hand can be constructed to bind any protein and ligand for which an aptamer exists.

small 2005, 1, No. 3, 284 –299 www.small-journal.com 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 295 reviews F. C. Simmel and W. U. Dittmer

6. DNA Devices and Information Processing A different approach towards autonomous computing with DNA is based on DNA enzymes (deoxyribozymes, see The seminal paper by Adleman[115] on the solution of Section 5.). Stojanovic and co-workers[126] designed a variety the Hamiltonian path problem using molecular biology of logic gates based on the action of the RNA-cleaving de- techniques marks the starting point for the fascinating re- oxyribozymes E6[127] and 8–17.[106,128] A chimeric DNA/RNA search area of DNA computing (see, for example, hybrid structure with an RNA base at the appropriate posi- Refs. [34,116–121]). An overview of this field will not be at- tion is cleaved by the DNA enzymes when bound to their tempted here, however, it should be stressed that it is impor- substrate recognition sections. These recognition sequences tant to also consider the information-carrying nature of can be protected by self-hybridized stem–loop modules. By DNA in combination with the information-processing abili- base-pairing with a DNA strand complementary to the loop ties of many enzymes as important parameters for the con- sequence, the stem–loop modules can be opened. This struction of future DNA-based nanodevices. makes the substrate recognition sequence accessible for the substrate to be cleaved. By combining several such control- ling and catalytic modules, it was possible to realize simple 6.1. Autonomous Computing with DNA logic gates such as AND or XOR,[126] a DNA half-adder,[129] and a molecular computer that could autonomously play An ingenious approach towards autonomous informa- (and win) the game “tic-tac-toe”.[130] tion processing with DNA-based devices was taken by Ben- The significance of these results for the field of DNA enson and co-workers.[122–124] They utilized the unique DNA- nanodevices is the fact that DNA molecules here serve as cleaving properties of the class IIS restriction endonuclease an input for an autonomous (molecular) computation proc- FokI in combination with clever DNA-sequence design to ess and also as an output. This opens up a fascinating possi- build the molecular realization of “finite automata”. Finite bility: Coupling a computational module to a DNA nano- automata are computing machines with a finite number of mechanical device would allow the control of the action of internal states, which can undergo transitions between these the device in response to an external molecular signal. This states according to certain rules. A computer program for could also be made conditional, that is, depending on the such an automaton prescribes a series of such transitions. nature of the signal, a variety of actions could be taken. The hardware for the automata originally consisted of the Similarly, the temporal/logical order of the action of a varie- restriction enzyme FokI, T4 ligase, and ATP (later it was ty of devices could be controlled by computational process- shown that ligase and ATP were not essential for the com- es. It is also conceivable that DNA-based devices could in- putation[123]). The software was a mixture of specially de- teract with each other by exchanging signaling molecules. signed DNA molecules. The operation of the automaton is based on the fact that FokI cleaves its double-stranded substrate 9 and 13 nt offset from its recognition site, thereby 6.2. Genetic Control Circuits creating a 4 nt sticky end. The various states and input sym- bols for the automaton are therefore represented by differ- It has long been recognized that biochemical reaction ent 4 nt long sequences. Benenson et al. use FokI to sequen- networks have information-processing properties.[131,132] This tially cleave off pieces of a double-stranded program and is particularly obvious in the case of genetic networks in thereby create a series of transitions between the various which the expression level of proteins is regulated by com- states of the automaton.[122] Finite automata are related to plex interactions between regulatory molecules (such as Turing machines but have limited computing power. The transcription factors) and genes. The best-studied example DNA automata were used to make simple decisions, such as of a genetic switch is the lac operon in E. coli bacteria, in to determine whether a given input symbol occurs more which the expression of the protein b-galactosidase is regu- than once. In a more recent paper, Benenson et al. were lated by the relative abundance of lactose and glucose. able to demonstrate how such DNA automata could actual- Since the discovery of the lac operon many gene regulatory ly find applications in autonomous diagnosis and drug-deliv- motifs have been found, which usually rely on feedback ery systems.[125] To this end, the transition molecules for a mechanisms and nonlinear cooperative phenomena. Quite FokI-based DNA automaton similar to the one described in recently, knowledge about these motifs has also been used Ref. [122] were designed to be active only in the presence to construct artificial gene networks. By transferring circuit of specific mRNA molecules. With these transition mole- designs known from electrical engineering to artificial regu- cules, a molecular computer could be realized that could latory networks, such functions as a bistable genetic decide whether the levels of certain mRNA molecules were switch,[133] a genetic oscillator,[134] or a sender–receiver high or low. Depending on the decision, a short DNA system[135] could be synthetically implemented in bacteria strand was administered as a “drug”. The authors chose a and similarly in vitro.[136] An overview of the developments realistic scenario: The mRNA levels sensed by their autom- in “synthetic biology” can be read elsewhere.[137] aton indicate the underexpression or overexpression of It is quite natural to think about a possible utilization of genes involved in prostate cancer. The DNA drug released gene regulatory mechanisms to control the behavior of after the computation could inhibit the synthesis of a DNA-based nanodevices as conformational changes of cancer-related protein by binding to its corresponding these devices can be brought about by DNA or RNA mRNA. strands. A first step in this direction was taken when DNA

296 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim www.small-journal.com small 2005, 1, No. 3, 284 –299 DNA Nanodevices tweezers (see Section 3.2.) were operated by an mRNA fuel questions, advanced biophysical techniques will have to be strand that was transcribed from an artificial “fuel gene” applied to these artificial molecular devices, in particular (see Figure 14).[138] By regulating the transcription of fuel techniques for single-molecule characterization. genes by using regulatory motifs known from natural and artificial genetic networks, it should be possible to program complex behavior into DNA devices or to intimately couple Acknowledgements their operation to genetic mechanisms. The authors wish to thank Jçrg P. Kotthaus for his support and Tim Liedl for critically reading the manuscript. Financial support by the Deutsche Forschungsgemeinschaft (Emmy Noether grant DFG SI 761/2–2) and the Alexander von Hum- boldt foundation is gratefully acknowledged.

[1] B. Alberts, D. Bray, J. Lewis, M. Raff, K. Roberts, J. D. Watson, Molecular Biology of the Cell, 3rd ed., Garland Publishing, New York, 1994. [2] V. A. Bloomfield, D. M. Crothers, I. Tinoco, Jr., Nucleic Acids, University Science Books, Sausalito, 2000. [3] R. H. Austin, J. P. Brody, E. C. Cox, T. Duke, W. Volkmuth, Phys. Today 1997, 50, 32–38. Figure 14. Transcriptional control of DNA tweezers: Instead of using a [4] W. M. Gelbart, R. F. Bruinsma, P. A. Pincus, V. A. Parsegian, DNA fuel strand, the fuel for the tweezers is encoded in an “artificial Phys. Today 2000, 53, 38–44. gene” with a promoter and a termination sequence for the T7 RNA [5] J. SantaLucia, Proc. Natl. Acad. Sci. USA 1998, 95, 1460– polymerase. The polymerase translates the fuel gene into “fuel 1465. mRNA” which closes the tweezers (adapted from Ref. [138]). [6] J. SantaLucia, D. Hicks, Annu. Rev. Biophys. Biomol. Struct. 2004, 33, 415–440. 7. Outlook [7] N. Peyret, P. A. Seneviratne, H. T. Allawi, J. SantaLucia, Bio- chemistry 1999, 38, 3468–3477. [8] N. Peyret, J. SantaLucia, HYTHER, version 1.0, http://ozone2.- In this Review, we have presented an overview of the chem.wayne.edu/, Wayne State University. current state of research into DNA-based devices. In addi- [9] B. Essevaz-Roulet, U. Bockelmann, F. Heslot, Proc. Natl. Acad. tion to the impressive accomplishments in DNA nanocon- Sci. USA 1997, 94, 11935–11940. struction, DNA has now proven to be also an interesting [10] U. Bockelmann, B. EssevazRoulet, F. Heslot, Phys. Rev. Lett. material for the realization of artificial molecular machines. 1997, 79, 4489–4492. DNA has been used to induce simple nanoscale movements, [11] M. Rief, H. Clausen-Schaumann, H. E. Gaub, Nat. Struct. Biol. 1999, 6, 346–349. to grab or release molecules, and to process information; it [12] M. Doi, S. F. Edwards, The Theory of Polymer Dynamics, Oxford has been further shown that these tasks can be performed in University Press, Oxford, 1986. an autonomous way. In the future, the various capabilities [13] J. Howard, Mechanics of Motor Proteins and the Cytoskeleton, of DNA devices will be integrated into more complex mo- 1st ed., Sinauer, Sunderland, 2001. lecular machines. One could think of a combination of a [14] S. B. Smith, L. Finzi, C. Bustamante, Science 1992, 258, DNA molecular computer, a DNA motor, and a DNA ap- 1122–1126. [15] S. B. Smith, Y. J. Cui, C. Bustamante, 1996, , 795– tamer in which the computer decides where a molecule at- Science 271 799. tached to a DNA aptamer is to be brought by the motor [16] M. D. Wang, H. Yin, R. Landick, J. Gelles, S. M. Block, Biophys. and when it has to be released. As shown above, all of the J. 1997, 72, 1335–1346. components for such a device have been demonstrated al- [17] B. Tinland, A. Pluen, J. Sturm, G. Weill, Macromolecules 1997, ready. Such systems could be important in the context of 30, 5763–5765. nanoassembly or intelligent drug delivery—depending on [18] C. Rivetti, C. Walker, C. Bustamante, J. Mol. Biol. 1998, 280, the application in an artificial or in a biomedical context. 41–59. [19] J. B. Mills, E. Vacano, P. J. Hagerman, J. Mol. Biol. 1999, 285, It will also be extremely interesting to see whether one 245–257. can couple these “semi-biological” devices to inorganic [20] F. Oosawa, Biopolymers 1968, 6, 1633–1647. nanostructures or, at the other extreme, to utilize them in a [21] G. S. Manning, J. Chem. Phys. 1969, 51, 924–933. biological context. Nano–bio hybrids could find applications [22] N. Gronbech-Jensen, R. J. Mashl, R. F. Bruinsma, W. M. Gelbart, as novel biosensors or biomolecular actuators; DNA devices Phys. Rev. Lett. 1997, 78, 2477–2480. could assist in molecular self-assembly or be part of biomo- [23] A. Y. Grosberg, T. T. Nguyen, B. I. Shklovskii, Rev. Mod. Phys. lecular diagnosis and drug-delivery systems. 2002, 74, 329–345. [24] C. M. Niemeyer, Angew. Chem. 1997, 109, 603–605; Angew. Of course, there are also many fundamental issues to be Chem. Int. Ed. Engl. 1997, 36, 585–587. addressed: What forces can be effectively generated by [25] N. C. Seeman, Nature 2003, 421, 427–431. DNA-based motors? What is the stalling force when under [26] N. C. Seeman, Sci. Am. 2004, 290(June), 34–43. load? How fast can a DNA motor run? To answer these [27] J. H. Chen, N. C. Seeman, Nature 1991, 350, 631–633. small 2005, 1, No. 3, 284 –299 www.small-journal.com 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 297 reviews F. C. Simmel and W. U. Dittmer

[28] Y. W. Zhang, N. C. Seeman, J. Am. Chem. Soc. 1994, 116, [61] C. M. Niemeyer, M. Adler, S. Gao, L. F. Chi, Angew. Chem. 1661–1669. 2000, 112, 3183–3187; Angew. Chem. Int. Ed. 2000, 39, [29] C. D. Mao, W. Q. Sun, N. C. Seeman, Nature 1997, 386, 137– 3055–3059. 138. [62] X. Li, D. R. Liu, Angew. Chem. 2004, 116, 4956–4978; Angew. [30] M. L. Petrillo, C. J. Newton, R. P. Cunningham, R. I. Ma, N. R. Chem. Int. Ed. 2004, 43, 4848–4870. Kallenbach, N. C. Seeman, Biopolymers 1988, 27, 1337– [63] C. D. Mao, W. Q. Sun, Z. Y. Shen, N. C. Seeman, Nature 1999, 1352. 397, 144–146. [31] A. I. H. Murchie, R. M. Clegg, E. von Kitzing, D. R. Duckett, S. [64] T. Fçrster, Ann. Phys. 1948, 2, 55–75. Diekmann, D. M. J. Lilley, Nature 1989, 341, 763–766. [65] L. Stryer, R. P. Haugland, Proc. Natl. Acad. Sci. USA 1967, 58, [32] P. S. Eis, D. P. Millar, Biochemistry 1993, 32, 13852–13860. 719–726. [33] T. J. Fu, N. C. Seeman, Biochemistry 1993, 32, 3211–3220. [66] J. R. Lakowicz, Principles of Fluorescence Spectroscopy, 2nd [34] E. Winfree, F. R. Liu, L. A. Wenzler, N. C. Seeman, Nature 1998, ed., Kluwer Academic, New York, 1999. 394, 539–544. [67] C. M. Niemeyer, M. Adler, S. Lenhert, S. Gao, H. Fuchs, L. F. [35] P. Sa-Ardyen, A. V. Vologodskii, N. C. Seeman, Biophys. J. 2003, Chi, ChemBioChem 2001, 2, 260–264. 84, 3829–3837. [68] R. P. Fahlman, M. Hsing, C. S. Sporer-Tuhten, D. Sen, Nano Lett. [36] T. H. LaBean, H. Yan, J. Kopatsch, F. R. Liu, E. Winfree, J. H. Reif, 2003, 3, 1073–1078. N. C. Seeman, J. Am. Chem. Soc. 2000, 122, 1848–1860. [69] R. H. Shafer, Prog. Nucleic Acid Res. Mol. Biol. 1998, 59,55– [37] H. Yan, S. H. Park, G. Finkelstein, J. H. Reif, T. H. LaBean, Sci- 94. ence 2003, 301, 1882–1884. [70] D. S. Liu, S. Balasubramanian, Angew. Chem. 2003, 115, [38] D. Liu, S. H. Park, J. H. Reif, T. H. LaBean, Proc. Natl. Acad. Sci. 5912–5914; Angew. Chem. Int. Ed. 2003, 42, 5734–5736. USA 2004, 101, 717–722. [71] Y. Chen, S. H. Lee, C. D. Mao, Angew. Chem. 2004, 116, [39] D. Liu, M. S. Wang, Z. X. Deng, R. Walulu, C. D. Mao, J. Am. 5449–5452; Angew. Chem. Int. Ed. 2004, 43, 5335–5338. Chem. Soc. 2004, 126, 2324–2325. [72] K. Gehring, J. L. Leroy, M. Gueron, Nature 1993, 363, 561– [40] B. Q. Ding, R. J. Sha, N. C. Seeman, J. Am. Chem. Soc. 2004, 565. 126, 10230–10231. [73] J. L. Leroy, K. Gehring, A. Kettani, M. Gueron, Biochemistry [41] W. M. Shih, J. D. Quispe, G. F. Joyce, Nature 2004, 427, 618– 1993, 32, 6019–6031. 621. [74] J. L. Leroy, M. Gueron, J. L. Mergny, C. Helene, Nucleic Acids [42] L. Jaeger, N. B. Leontis, Angew. Chem. 2000, 112, 2576– Res. 1994, 22, 1600–1606. 2580; Angew. Chem. Int. Ed. 2000, 39, 2521–2524. [75] R. J. Sanyal, Masters thesis, Princeton University (USA), 2002. [43] L. Jaeger, E. Westhof, N. B. Leontis, Nucleic Acids Res. 2001, [76] B. Yurke, A. J. Turberfield, A. P. Mills, F. C. Simmel, J. L. Neu- 29, 455–463. mann, Nature 2000, 406, 605–608. [44] E. Braun, Y. Eichen, U. Sivan, G. Ben-Yoseph, Nature 1998, [77] C. Green, C. Tibbetts, Nucleic Acids Res. 1981, 9, 1905–1918. 391, 775–778. [78] I. G. Panyutin, P. Hsieh, Proc. Natl. Acad. Sci. USA 1994, 91, [45] J. Richter, R. Seidel, R. Kirsch, M. Mertig, W. Pompe, J. 2021–2025. Plaschke, H. K. Schackert, Adv. Mater. 2000, 12, 507–510. [79] F. C. Simmel, B. Yurke in Smart Structures And Materials 2001: [46] J. Richter, M. Mertig, W. Pompe, I. Monch, H. K. Schackert, Industrial And Commerical Applications Of Smart Structures Appl. Phys. Lett. 2001, 78, 536–538. Technologies, Vol. 4332, 2001, pp. 419–428. [47] J. Richter, M. Mertig, W. Pompe, H. Vinzelberg, Appl. Phys. A [80] F. C. Simmel, B. Yurke, R. J. Sanyal, J. Nanosci. Nanotechnol. 2002, 74, 725–728. 2002, 2, 383–390. [48] K. Keren, M. Krueger, R. Gilad, G. Ben-Yoseph, U. Sivan, E. [81] F. C. Simmel, B. Yurke, Phys. Rev. E 2001, 63, 041913. Braun, Science 2002, 297, 72–75. [82] F. C. Simmel, B. Yurke, Appl. Phys. Lett. 2002, 80, 883–885. [49] J. Richter, M. Adler, C. M. Niemeyer, ChemPhysChem 2003, 4, [83] J. C. Mitchell, B. Yurke in DNA computing—7th International 79–83. Workshop on DNA-Based Computers, DNA 7, Vol. 2340 (Eds.: [50] J. L. Coffer, S. R. Bigham, X. Li, R. F. Pinizzotto, Y. G. Rho, R. M. N. Jonoska, N. C. Seeman), Springer, Heidelberg, 2002, Pirtle, I. L. Pirtle, Appl. Phys. Lett. 1996, 69, 3851–3853. pp. 258–268. [51] W. U. Dittmer, F. C. Simmel, Appl. Phys. Lett. 2004, 85, 633– [84] H. Yan, X. P. Zhang, Z. Y. Shen, N. C. Seeman, Nature 2002, 635. 415, 62–65. [52] R. Nagarajan, W. Liu, J. Kumar, S. K. Tripathy, F. F. Bruno, L. A. [85] Z. Y. Shen, H. Yan, T. Wang, N. C. Seeman, J. Am. Chem. Soc. Samuelson, Macromolecules 2001, 34, 3921–3927. 2004, 126, 1666–1674. [53] N. Kobayashi, S. Uemura, K. Kusabuka, T. Nakahira, H. Takaha- [86] L. P. Feng, S. H. Park, J. H. Reif, H. Yan, Angew. Chem. 2003, shi, J. Mater. Chem. 2001, 11, 1766–1768. 115, 4478–4482; Angew. Chem. Int. Ed. 2003, 42, 4342– [54] Y. F. Ma, J. M. Zhang, G. J. Zhang, H. X. He, J. Am. Chem. Soc. 4346. 2004, 126, 7097–7101. [87] J. W. J. Li, W. H. Tan, Nano Lett. 2002, 2, 315–318. [55] K. A. Williams, P. T. M. Veenhuizen, B. G. de la Torre, R. Eritja, C. [88] P. Alberti, J. L. Mergny, Proc. Natl. Acad. Sci. USA 2003, 100, Dekker, Nature 2002, 420, 761. 1569–1573. [56] K. Keren, R. S. Berman, E. Buchstab, U. Sivan, E. Braun, Sci- [89] J. S. Shin, N. A. Pierce, Nano Lett. 2004, 4, 905–909. ence 2003, 302, 1380–1382. [90] A. J. Turberfield, J. C. Mitchell, B. Yurke, A. P. Mills, M. I. Blakey, [57] C. A. Mirkin, R. L. Letsinger, R. C. Mucic, J. J. Storhoff, Nature F. C. Simmel, Phys. Rev. Lett. 2003, 90, 118102. 1996, 382, 607–609. [91] J. T. Finer, R. M. Simmons, J. A. Spudich, Nature 1994, 368, [58] A. P. Alivisatos, K. P. Johnsson, X. G. Peng, T. E. Wilson, C. J. 113–119. Loweth, M. P. Bruchez, P. G. Schultz, Nature 1996, 382, 609– [92] K. Svoboda, C. F. Schmidt, B. J. Schnapp, S. M. Block, Nature 611. 1993, 365, 721–727. [59] G. P. Mitchell, C. A. Mirkin, R. L. Letsinger, J. Am. Chem. Soc. [93] S. C. Kuo, M. P. Sheetz, Science 1993, 260, 232–234. 1999, 121, 8122–8123. [94] D. Bray, Cell Movements, 2nd ed., Garland Publishing, New [60] S. J. Park, A. A. Lazarides, C. A. Mirkin, P. W. Brazis, C. R. Kan- York, 2001. newurf, R. L. Letsinger, Angew. Chem. 2000, 112, 4003– 4006; Angew. Chem. Int. Ed. 2000, 39, 3845–3848.

298 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim www.small-journal.com small 2005, 1, No. 3, 284 –299 DNA Nanodevices

[95] P. Yin, H. Yan, X. G. Daniell, A. J. Turberfield, J. H. Reif, Angew. [119] A. P. Mills, B. Yurke, P. M. Platzman, Biosystems 1999, 52, Chem. 2004, 116, 5014–5019; Angew. Chem. Int. Ed. 2004, 175–180. 43, 4906–4911. [120] B. Yurke, A. P. Mills, S. L. Cheng, Biosystems 1999, 52, 165– [96] J. S. Shin, N. A. Pierce, J. Am. Chem. Soc. 2004, 126, 10834– 174. 10835. [121] D. Faulhammer, A. R. Cukras, R. J. Lipton, L. F. Landweber, Proc. [97] W. B. Sherman, N. C. Seeman, Nano Lett. 2004, 4, 1203– Natl. Acad. Sci. USA 2000, 97, 1385–1389. 1207. [122] Y. Benenson, T. Paz-Elizur, R. Adar, E. Keinan, Z. Livneh, E. Sha- [98] Y. Tian, C. D. Mao, J. Am. Chem. Soc. 2004, 126, 11410– piro, Nature 2001, 414, 430–434. 11411. [123] Y. Benenson, R. Adar, T. Paz-Elizur, Z. Livneh, E. Shapiro, Proc. [99] D. S. Wilson, J. W. Szostak, Annu. Rev. Biochem. 1999, 68, Natl. Acad. Sci. USA 2003, 100, 2191–2196. 611–647. [124] R. Adar, Y. Benenson, G. Linshiz, A. Rosner, N. Tishby, E. Sha- [100] E. A. Doherty, J. A. Doudna, Annu. Rev. Biochem. 2000, 69, piro, Proc. Natl. Acad. Sci. USA 2004, 101, 9960–9965. 597–615. [125] Y. Benenson, B. Gil, U. Ben-Dor, R. Adar, E. Shapiro, Nature [101] G. F. Joyce, Annu. Rev. Biochem. 2004, 73, 791–836. 2004, 429, 423–429. [102] A. D. Ellington, J. W. Szostak, Nature 1990, 346, 818–822. [126] M. N. Stojanovic, T. E. Mitchell, D. Stefanovic, J. Am. Chem. [103] C. Tuerk, L. Gold, Science 1990, 249, 505–510. Soc. 2002, 124, 3555–3561. [104] L. C. Bock, L. C. Griffin, J. A. Latham, E. H. Vermaas, J. J. Toole, [127] R. R. Breaker, G. F. Joyce, Chem. Biol. 1995, 2, 655–660. Nature 1992, 355, 564–566. [128] J. Li, W. C. Zheng, A. H. Kwon, Y. Lu, Nucleic Acids Res. 2000, [105] D. L. Robertson, G. F. Joyce, Nature 1990, 344, 467–468. 28, 481–488. [106] S. W. Santoro, G. F. Joyce, Proc. Natl. Acad. Sci. USA 1997, 94, [129] M. N. Stojanovic, D. Stefanovic, J. Am. Chem. Soc. 2003, 125, 4262–4266. 6673–6676. [107] N. Carmi, S. R. Balkhi, R. R. Breaker, Proc. Natl. Acad. Sci. USA [130] M. N. Stojanovic, D. Stefanovic, Nat. Biotechnol. 2003, 21, 1998, 95, 2233–2237. 1069–1074. [108] D. P. Bartel, J. W. Szostak, Science 1993, 261, 1411–1418. [131] A. Arkin, J. Ross, Biophys. J. 1994, 67, 560–578. [109] E. H. Ekland, J. W. Szostak, D. P. Bartel, Science 1995, 269, [132] D. M. Wolf, A. P. Arkin, Curr. Opin. Microbiol. 2003, 6, 125– 364–370. 134. [110] B. L. Zhang, T. R. Cech, Nature 1997, 390, 96–100. [133] T. S. Gardner, C. R. Cantor, J. J. Collins, Nature 2000, 403, [111] C. M. Niemeyer, M. Adler, Angew. Chem. 2002, 114, 3933– 339–342. 3937; Angew. Chem. Int. Ed. 2002, 41, 3779–3783. [134] M. B. Elowitz, S. Leibler, Nature 2000, 403, 335–338. [112] W. U. Dittmer, A. Reuter, F. C. Simmel, Angew. Chem. 2004, [135] R. Weiss, T. F. Knight, Jr. in DNA Computing, 6th International 116, 3634–3637; Angew. Chem. Int. Ed. 2004, 43, 3550– Workshop on DNA-Based Computers, DNA6, Vol. 2054 (Eds.: 3553. A. E. Condon, G. Rozenberg), Springer, Heidelberg, 2000, [113] Y. Chen, M. S. Wang, C. D. Mao, Angew. Chem. 2004, 116, pp. 1–16. 3638–3641; Angew. Chem. Int. Ed. 2004, 43, 3554–3557. [136] V. Noireaux, R. Bar-Ziv, A. Libchaber, Proc. Natl. Acad. Sci. USA [114] Y. Chen, C. D. Mao, J. Am. Chem. Soc. 2004, 126, 8626–8627. 2003, 100, 12672–12677. [115] L. M. Adleman, Science 1994, 266, 1021–1024. [137] J. Hasty, D. McMillen, J. J. Collins, Nature 2002, 420, 224– [116] R. J. Lipton, Science 1995, 268, 542–545. 230. [117] F. Guarnieri, M. Fliss, C. Bancroft, Science 1996, 273, 220– [138] W. U. Dittmer, F. C. Simmel, Nano Lett. 2004, 4, 689–691. 223. [118] Q. Ouyang, P. D. Kaplan, S. M. Liu, A. Libchaber, Science 1997, Received: October 29, 2004 278, 446–449.

small 2005, 1, No. 3, 284 –299 www.small-journal.com 2005 Wiley-VCH Verlag GmbH & Co. KGaA, D-69451 Weinheim 299